CAPÍTULO IV. DESCRIPCIÓN DEL CONTEXTO
1. ANALISIS DE LAS CARACTERISTICAS DEL ENTORNO ESCOLAR
Figure 5.3 Comparing individual tracker profiles
Some of our initial collaborative outputs are presented below. Figure 5.3 shows a profile of particular tracking toolkits, listing the different kinds of data that they can collect. Each component in the toolkit stands for the collection of different data types. For example, in the visualisation of the Google+ widget, the buttons on the top row stand for ‘Ad Views’ (AV), ‘Analytics’ (A), and ‘Browser Information’ (BI). The second row contains buttons representing ‘Cookie Data’ (CO), ‘Date/Time’ (D/T) and ‘Demographic Data’ (DD). The measuring tape stands for the collection of phone numbers and the black knitting needles towards the bottom-right stand for ‘Device ID’ (DI). These visualisations tell us that in order for companies to engage in online profiling, trackers need to stitch cookies into your browser, pin down your device ID, and obtain browser information and information about the date and time of a visit. Through these methods trackers can weave data together into more or less personal profiles. In order to give some indication of how personal this gets, the legend on the left side of the image shows three different symbols, each of which represents a sort of data: Anonymous, Pseudonymous and Personally Identifiable Data (PII) (these draw on Ghostery’s own categorisations).
One of the results of these trackers-as-toolkits visualisations is that, when placed next to one another, they render more immediately visible the differing levels of sophistication possess by different trackers; here, for instance, how the Google+ tracker deploys a far greater variety of different ways to collect data than Clicktale.
This profiling of individual trackers has generic applicability. More specific to our research object is the profiling of how different digital subprime sites bring different trackers together. Three examples are shown (Figure 5.4). Each website is represented by a section of fabric, composed by all the trackers active in that website. The vertical length of this fabric indicates the number of trackers it contains. Each tracker, in turn, is identified with a number of threads proportional to the number of different types of data tracked by that tracker. This means that the ‘density’ of threads, the degree to which they are entangled, also allows for easy comparison between trackers. The design of this visualisation is quite deliberate: the curly/tortuous style of the threads is intended to give the images a sense of instability, signalling that the trackers and the collected data types may vary over time and over different browsing sessions. A dashed line indicates when no information is disclosed about what data a particular tracker collects. The icons to the left of the chart lines summarise the stated data retention policies of each tracker. For some trackers this is 18–24 months, for others it is a few years. More often than not, this remains undisclosed. Lastly, the icons on the right represent each tracker’s ‘data sharing’ policy. It indicates what kind of data – for instance, aggregate data, anonymous data, and PII data – is shared with third parties.21
Comparing the ‘tracker profile’ of Kredito24.es (Spain, owned by Kreditech), Wonga (UK) and Spotloan’s (USA, using technology licensed from ZestFinance (Hardy 2012)), we begin to be able to better detect important points of commonality and difference. Kredito24.es outweighs the other two in terms of the density of the data being tracked. All three are heavily reliant on trackers that do not disclose their data retention period.22
They are also reliant on trackers that provide anonymous information about a particular user to third parties. For digital subprime lenders, what is important is the creation of specific profiles about their visitors in order to aid credit assessment. When combined with personalised data input by a potential borrower, this anonymous information can be tied to the individual. We can also see that Wonga is using a number of trackers whose data collection functions remain opaque – at least to Ghostery.
Figure 5.4 Comparing Kredito24.es, Wonga and Spotloan
Further, the creation of individual profiles points us to what is unique in the tracking fabrics being composed by different digital subprime sites. In Wonga’s case, a unique tracker is QuBit OpenTag. QuBit is a London based tech company, funded, perhaps
coincidentally, by Balderton Capital, the same venture capitalist firm as Wonga. OpenTag
itself is a tool partly designed to help companies improve their website’s performance and monitoring. But QuBit also helps websites provide exactly the kind of real time personalised content, based on data such as browser type and IP address, that make Wonga’s slider appear at different initial positions for different people. Thus, in a report designed to showcase the power of their analytics, QuBit describes how technology purchases by visitors using Safari are “around £30 more than any other browser”, a conclusion designed to assist in practices of customer segmentation (Qubit, 2013: 14). Crucially this can be done virtually instantly, based on variables that many users might assume to be irrelevant.
Spotloan, meanwhile, is unique in using a tracker called ‘ThreatMetrix’. In an industry
sales briefing, ThreatMetrix is described as a provider of integrated cybercrime prevention
solutions. The ThreatMetrix™ Cybercrime Defender Platform helps companies protect
man-in-the browser (MitB) and Trojan attacks (Threatmetrix, 2013: 2).
The tracker is significant because it shows how the industry of online tracking, often associated with understanding and shaping consumer practices, is in this instance linking up with an industry concerned with combating cybercrime. Establishing the identity of a borrower has long been central to credit assessment practices. As these practices move online, new opportunities for potential fraudsters open up, thus generating new challenges for creditors. Such trackers are an indication of tailored attempts to manage the emergent risks involved.
Conclusion
Our attempts to understand the algorithmic basis of digital subprime calculation are ongoing. Since this initial pilot project we have expanded our dataset to incorporate twenty sites we suspect of using similar techniques and are now collecting data on a monthly basis. We also aim to further engage with industry figures. We are thus still in the process of chipping away at the opacities that characterise this industry. Digital methods are tools to do so, but they will need to be combined with others. In order to understand the rise and significance of so called ‘big data’ analytics, we as researchers will thus likely have to rely on a diverse palette of approaches, not just to keep our objects stable and ‘detectable’ (Law, 2009), but also to be able to understand and to become attuned to their transformations as they pass through diverse of socio-technical registers.
Our initial research has, however, provided both insights into the tracking work being done by digital subprime trackers, as well as into the challenges facing researchers seeking to understand online algorithmic calculation from the outside of an industry. In respect of the former, we can return to Ossandon (2013), who suggests that, while we know that credit
practices produce networks, “what kind of collective or social formation are we talking about? At what level do these networks operate?” In the case of digital subprime, our initial findings suggest the creation of networks not just between potential borrowers and organisations involved in the credit industry (including both lenders and third party credit reference agencies), but also now involving the ever growing industry of online tracking. These sites do, then, seem to have an interest in using trackers to collect user data for the purposes of credit assessment and online behavioural profiling and segmentation. Consumer credit lending has long been accompanied by a range of controversies (see: Deville, 2015). In the case of digital subprime, there is the potential for it to become wrapped up in the controversies surrounding the ethics of online tracking and the collection and retention of the data of users. Further, the deployment of ‘custom’ trackers and the common interest in particular data types also suggests an industry-specific ‘professionalisation’ of tracking practices. In other words, this is the highly emergent, likely experimental deployment of trackers that meet the specific needs of digital subprime websites.
Finally, we can reflect on what kinds of transparency such methods produce. Our findings are, to a degree, achievements of transparency, even if they remain incomplete. Striving to open up the opacities of digital subprime has also pointed us to the way in which digital methods itself is involved in the production of opacity. In our case, this has centred most clearly on our dependence on Ghostery’s database and its process of
categorisation. We have, however, departed from Ghostery’s elementary understanding of
trackers, and moved to a vocabulary of threads and density, which we consider more appealing to describe unseen ‘work’ of trackers (Star, 1991). Further, the role of rendering
visual what is usually unseen is also centrally important to our work. As Tyler Reigeluth (2014) notes, digital traces tend to be naturalised and claims can too readily be made about their objectivity. He proposes to see such traces as “in-formation” (Reigeluth, 2014: 253). For our emergent sociology of the invisible, the challenge has been, and continues to be, to grasp how trackers partake in forming digital traces and how they are also traces in formation themselves. One way we have begun to grapple with these issues is through visualisations that have emerged as the product of collaboration with designers. These reflections have helped us in turn to profile the different digital subprime websites, as different kinds and unstable textures. The challenge as we take this project forward is how to track and render visible these textures, as they continue to be re-shaped and knitted anew.
Acknowledgements
The authors would like to thank the editors and Paul Langley for their helpful comments on a previous draft of this chapter. Many thanks also to Frederica Bardelli and Carlo de Gaetano for their hard work on the visualisations and to Erik Borra and Emile den Tex for their assistance with the Tracker Tracker tool.
Notes
1 The way Star analyses ‘social control’ does not map cleanly into our case. Star is, amongst other things, concerned with the unseen, unrecognised and unpaid workers, such as (the many women) involved in unpaid (home) care. Their becoming invisible goes hand in hand with quite explicit forms of social control. Although the machines we study are indeed undertaking invisible work, their invisibility would connect up to very different forms of control – and would have less to do with their suppression as labouring subjects.
2 Although less relevant for present purposes, we are also interested and include under the ‘digital subprime’ umbrella another set of ventures that use alternative online methods to try to assess user behaviour (e.g., Lenddo and LendUp, both based in the US).
3 To be clear, these are individuals to whom the status ‘subprime’ is assigned by the credit industry, rather than a particular quality of personhood. Although the term is formally used within the industry to refer to borrowers who have fallen below a precise threshold in a risk-based analysis of creditworthiness, it is also used to refer more generally to those categories of borrowers that are perceived, irrespective of any formal evaluation, as undesirable from the point of view of mainstream lenders (see Langley, 2008: 473).
4 A site possibly using similar techniques is Sunny (www.sunny.co.uk). It is run by Think Finance, known for its use of big data analytics.
5 The ‘mainstream’/‘non-mainstream’ distinction is a placeholder used for convenience. Payday lending is very much on a spectrum of credit products available to potential borrowers, and it would be incorrect to label it as in any way a separate domain. This is in particular given both the quantitative increase in such businesses in many countries, including the UK, and the fact that this industry relies extensively on ostensible ‘mainstream’ credit scoring technologies. It might better, therefore, be considered an example of what Rob Aitken (2006, 2010) calls ‘fringe finance’ (see also: Langley (2008: 170)), given that the metaphor evokes a continuity.
6 In the UK, an individual may have a number of such ratings, each generated by competing credit reference agencies. Wonga, for instance, draws data from both Callcredit and Experian (Wonga, 2014c).
7 At present it is unclear exactly what this additional third party data comprises (which is unsurprising, given that the information is proprietary). One source in a recent article in The Guardian newspaper speculates that Wonga draws on the wealth of free information that is available instantly online: electoral roll details, estimates of house values, for instance (Lewis, 2011). Wonga asks for users’ vehicle registration details in the application process (if they own a vehicle). This might suggest they are tapping into the database of registered owners, perhaps to verify identity, perhaps to feed this into their risk calculations. Another source we spoke to speculated that they might also look into databases containing stolen mobile numbers (users are also asked to provide their mobile number), which could, again, be used to feed into their risk calculations.
8 Discussion with an anonymous industry source, 28 October 2013.
9 See, for instance, research into public views on targeted advertising (Pew Research Center, 2012), discussions about ‘Do Not Track’ (www.eff.org/issues/do-not-track), which dates from 2007, but also the writings by public intellectuals, such as Evgeny Morozov, about consumer surveillance (and about ‘big data’ in relation to credit assessments) (Morozov, 2013).
10 See for instance, Lightbeam (www.mozilla.org/en-US/lightbeam/) and Disconnect (https://disconnect.me/).
11 For example, through moving the nodes of your online ‘data body’ (Lightbeam), testing the effect of blocking trackers to your browsing experience such as connection speed (Disconnect), or by engaging with an analysis of rankings (Ghostery).
12 See: www.ghostery.com.
13Ghostery. ‘How It Works.’ www.ghostery.com/how-it-works (accessed on 20 January 2014).
14 Evidon. ‘Ghostery Sees What Scanners Alone Can’t.’ www.evidon.com/analytics (accessed on 8 March 2014).
15 See: http://knowyourelements.com (accessed 21 September 2014).
16 https://tools.digitalmethods.net/beta/trackerTracker/. The tool was created in a collaborative project by Yngvil Beyer, Erik Borra, Carolin Gerlitz, Anne Helmond, Koen Martens, Simeona Petkova, JC Plantin, Bernhard Rieder, Lonneke van der Velden, Esther Weltevrede at the Digital Methods Winter School 2012.
17 The Tracker Tracker found third party content on 72 per cent of the sites in the sample; using manual methods the figure was 73 per cent.
18 Project page: https://wiki.digitalmethods.net/Dmi/TrackersGuide.
19 See discussion here: https://twitter.com/Ghostery/status/433349897471799296.
20 For example, IP addresses could until recently only give an indication of geographical location and could not match the geodemographic precision of, say, a UK postcode (on which see Burrows and Gane (2006)), although this kind of research is progressing quickly (Lowenthal, 20 April 2011).
21 ‘Sharing PII data with third parties’ does not necessarily mean that data is shared with any third party, such as the digital subprime website itself. It could also include another company that a tracker collaborates with, or an advertising network or broker.
References
Aitken, R. (2006) “Capital at its fringes”. New Political Economy, 11(4): 479–498.
Aitken, R. (2010) “Regul(ariz)ation of fringe credit: payday lending and the borders of global financial practice”. Competition & Change, 14(2): 80–99.
Bollier, D. (2010) The Promise and Peril of Big Data. Washington, DC: The Aspen Institute.
Borgesius, F.Z. (2013) “Behavioral targeting: a European legal perspective”. IEEE Security & Privacy, 11(1): 82–85.
Borgesius, F.Z. (2014) Behavioral Targeting: How to Regulate? (Diss.). Institute for
Information Law (IViR), University of Amsterdam.
Borra, E.K. and B. Rieder (2014) “Programmed method. Developing a toolset for capturing and analyzing Tweets”. In K. Weller and A. Bruns (eds). Aslib Journal of Information Management, 66(3): 262–278.
Burrows, R., and N. Gane (2006) “Geodemographics, software and class”. Sociology, 40(5):
793–812.
Castellucia, C.S. Grumbach, and L. Olejnik (2013) “Data harvesting 2.0: From the visible to the invisible web”. The Twelfth Workshop on the Economics of Information Security.
http://hal.inria.fr/hal-00832784 (last accessed 31 October 2014).
Chun, W. (2013) “Habitual New Media: Exposing Empowerment”, Barnard College, New
York City, 10 October 2013. http://vimeo.com/78287998 (last accessed 31 October 2014). Deville, J. (2012) “Regenerating market attachments: consumer credit debt collection and
the capture of affect”. Journal of Cultural Economy, 5(4): 423–439.
Deville, J. (2015). Lived Economies of Default: Consumer Credit, Debt Collection and the Capture of Affect. London: Routledge.
Gerlitz, C. and A. Helmond (2013) “The like economy: social buttons and the data-intensive Web”. New Media & Society, 15(8): 1348–1365.
Hardy, Q. (2012) “Big Data for the Poor”. The New York Times: Bits Blog.
http://bits.blogs.nytimes.com/2012/07/05/big-data-for-the-poor/ (last accessed 31 October 2014).
Helmond, A. (2013) “The algorithmization of the hyperlink.” Journal of Computational Culture, 3. http://computationalculture.net/article/the-algorithmization-of-the-
hyperlink (last accessed 31 October 2014).
Hoofnagle, C.J., Soltani, A., Good, N., Wambach, D.J., and M. Ayenson (2012) “Behavioral advertising: the offer you cannot refuse”. SSRN Scholarly Paper ID 2137601. Rochester,
accessed 31 October 2014).
Koot, M. (2012) “A survey of privacy and security decreasing third-party content on Dutch websites”, October. www.madison-gurkha.com/press/2012-10-
SurveyOfThirdPartyContent.pdf (last accessed 31 October 2014).
Kreditech (2013a) “Where We Are | Kreditech.com.” www.kreditech.com/ (last accessed 16 August 2015).
Kreditech (2013b) “Kreditechnology”. www.kreditech.com/ (last accessed 31 October 2014). Kreditech (2013c) “Online Loan and Online Lenders in Austria”. www.kreditech.com/ (last
accessed 31 October 2014).
Langley, P. (2008) “Sub-prime mortgage lending: a cultural economy.” Economy and Society,
37(4): 469–494.
Langley, P. (2014) “Equipping entrepreneurs: consuming credit and credit scores”.
Consumption Markets & Culture, 17(5): 448–467.
Law, J. (2009) “The materials of STS”, version 9 April 2009.
www.heterogeneities.net/publications/Law2008MaterialsofSTS.pdf (last accessed 31 October 2014).
Lewis, T. (2011) “With Wonga, your prosperity could count on an algorithm”, 25 October.
The Guardian. www.theguardian.co.uk/money/2011/oct/16/wonga-algorithm-lending-
debt-data (last accessed 31 October 2014).
Lowenthal, T. (2011) “IP address can now pin down your location to within a half mile”, April. Arstechnica. http://arstechnica.com/tech-policy/2011/04/getting-warmer-an-ip-
address-can-map-you-within-half-a-mile/ (last accessed 31 October 2014).
Marres, N. (2005) No Issue, No Public: Democratic Deficits After the Displacement of Politics. Amsterdam: Universiteit van Amsterdam.
Marres, N. (2012) “The redistribution of methods: on Intervention in digital social research, broadly conceived”. The Sociological Review, 60(S1): 139–165.
Marres, N. and E. Weltevrede. (2013) “Scraping the social? Issues in live social research”.
Journal of Cultural Economy, 6(3): 313–335.
Marron, D. (2009) Consumer Credit in the United States: A Sociological Perspective from the 19th Century to the Present. New York: Palgrave.
Mendick, R. (2012) “Justin Welby, the Bishop of Durham, Calls for Cap on Total Cost of Payday Loans”, 25 November, Telegraph.
www.telegraph.co.uk/news/religion/9700796/Justin-Welby-the-Bishop-of-Durham- calls-for-cap-on-total-cost-of-payday-loans.html (last accessed 16 August 2015). Morozov, E. (2013) “Your social networking credit score”, 30 January, Slate.
www.slate.com/articles/technology/future_tense/2013/01/wonga_lenddo_lendup_big_data_and_social_networking_banking.html
(last accessed 31 October 204).
Ossandon, J. (2012) “Sowing financial consumers: Retail credit in Chile”. In Isaconf.
https://isaconf.confex.com/isaconf/forum2012/webprogram/Paper21552.html (last accessed 31 October 2014).
Ossandon, J. (2013) “Low and high finance studies after performativity: A speculative workshop report, Charisma – Consumer Market Studies”, 8 July. www.charisma-
network.net/markets/low-and-high-finance-studies-after-performativity-a-speculative- workshop-report (last accessed 31 October 2014).
Pew Research Center (2012) “Search Engine Use 2012”. Washington: Pew Research Center’s
Internet and American Life Project, 9 March. www.pewinternet.org/2012/03/09/search-